GitHub Repository

Ultra-lightweight (24-byte RAM), real-time safety interceptor for RL agents and LLM control on embedded hardware. MISRA-C compliant, <1.2µs latency.

0 stars

MicroSafe-RL – Sub-microsecond safety layer for Edge AI 1.18µs latency

Name: MicroSafe-RL – Sub-microsecond safety layer for Edge AI 1.18µs latency
Availability: InStock
Author: DREDREG

by DREDREG·Apr 2, 2026·2 points·0 comments

Visit Project View on HN

AI Analysis

●●●BangerWizardryNiche GemBig Brain

1.18µs safety layer on STM32 beats software watchdogs by orders of magnitude.

Strengths

•Sub-microsecond latency (85 cycles at 72MHz) enables real-time hardware protection.
•Zero malloc and 20 bytes RAM makes it viable on cheapest microcontrollers.
•Python Auto-Tuner generates C++ parameters from 2 minutes of telemetry data.

Weaknesses

•Commercial license limits open-source adoption and community contributions.
•Extremely niche audience limits broader developer interest and feedback.

Post Description

I built MicroSafe-RL to solve the "Hardware Drift" problem in Reinforcement Learning. When RL agents move from simulation to real hardware, they often encounter unknown states and destroy expensive parts.

Key specs:

1.18µs latency (85 cycles on STM32 @ 72MHz)

20 bytes of RAM (no malloc)

Model-free: It adapts to mechanical wear-and-tear using EMA/MAD stats.

Includes a Python Auto-Tuner to generate C++ parameters from 2 mins of telemetry.

Check it out: https://github.com/Kretski/MicroSafe-RL